AI: July-born startup Sarvam raises $41 mn to work across multiple Indian languages

Virendra Pandit

New Delhi: Bengaluru-based, four-month-old startup Sarvam AI has raised USD 41 million (INR 3,417 crores), the largest such round by an early-stage Artificial Intelligence contender, in fresh funding to tap growth in the multilingual Indian market.

Sarvam AI is building large language models targeted at unique uses in Indian languages, aimed at making apps available at affordable price points, the media reported on Thursday. It had earlier secured USD 12 million as seed funding to train Indic large language modules (LLMs).

The new, homegrown startup with only 18 employees plans to use the funding to focus on India-specific offerings including training AI models to support the diverse set of Indian languages and voice-first interfaces. It aims to create population-scale impact by layering Gen AI on top of the highly successful India stack specifically for public goods applications.

Generative AI is a type of AI technology that can produce various types of content, including text, imagery, audio, and synthetic data, and conversational AI refers to technologies like chatbots or virtual agents to which users can talk.

The Series A funding round was led by Lightspeed Venture Partners, and participants included Silicon Valley billionaire Vinod Khosla’s Khosla Ventures and Peak XV Partners, formerly Sequoia Capital India & Southeast Asia, Sarvam said Thursday.

The nascent, relatively unknown startup was co-founded by Vivek Raghavan and Pratyush Kumar with plans to build open-source foundational AI systems as well as tools for creating apps for the meager compute infrastructure available to developers in India.

“What we are showing is that you can build large language models with limited resources,” Raghavan was quoted as saying. “We are building at a smaller scale and demonstrating that these can be extremely cost and energy-efficient so they can be accessed by everyone.”

Large language models, like OpenAI’s GPT-4 and Meta Platforms Inc.’s Llama, are powerful AI systems that understand by learning from vast amounts of diverse data from the internet and elsewhere to summarize, translate, and create text, audio, and video for a wide range of applications.

Sarvam’s first open-source model will work in the top 10 Indian languages and be released in the coming weeks. It will be available for trials by developers, startups, enterprises, and government-owned entities. Compared with the trillions of parameters that the largest models have been trained on, Sarvam’s systems will be “much smaller” and in the “billions,” Raghavan said.

The name Sarvam is derived from the ancient Sanskrit word meaning “All” to denote its inclusive mission, he said.

Raghavan is a computer engineer with a Ph.D. from Carnegie Mellon University, while Kumar is an alumnus of the Indian Institute of Technology-Bombay (IIT-B) with a computer engineering Ph.D. from ETH Zurich.

Indian-American Vinod Khosla, the co-founder of Sun Microsystems, a billionaire businessman, a Silicon Valley investment pioneer, and an early backer of OpenAI said several countries are driving sovereign efforts to build generative AI models given their strategic importance. “We need companies like Sarvam AI to develop deep expertise for building AI in and for India,” he said in a statement. 

The two co-founders of Sarvam previously worked at Infosys co-founder Nandan Nilekani-backed AI4Bharat. Sarvam is developing a full-stack offering for Generative AI, ranging from research-led innovations in training custom AI models to an enterprise-grade platform for authoring and deployment.

Raghavan, an entrepreneur and technologist instrumental in building Digital Public Goods (DPGs) like Aadhaar, said that Sarvam will work with Indian enterprises to co-develop domain-specific AI models leveraging their data.

The funding for Sarvam comes at a time when a host of investors are betting big on Generative AI startups in India.

The rapid adoption and rise in demand for generative AI tools like ChatGPT, has resulted in the rising number of Indian GenAI-based Software-as-a-Service (SaaS) startups which has more than doubled since 2021. The firms have raised more than USD 590 million in funding as of May 2023, according to a report by SaaSBoomi and McKinsey.

“Vivek’s expertise in building large-scale systems with Pratyush’s domain expertise in AI makes them uniquely positioned to build population-scale AI applications. Large-scale adoption of AI in India will require not only building uniquely Indian use cases but also delivering them at prices that everyone can afford and we believe the Sarvam team is best positioned to accomplish this,” said Harshjit Sethi, Managing Director, Peak XV.